Matthias Gonz´alez Benoˆıt Commer¸con, Neil Vaytet Rapha ...

Simulations of massive magnetized dense core collapse

Matthias Gonzalez

Benoıt Commercon, Neil VaytetRaphael Mignon-Risse

Laboratoire AIM, Universite Paris Diderot-CEA-CNRS, CEA Saclay

Outline

1 Context

2 Massive dense core collapse simulationsSetupMorphologiesOutflowDisc

3 Conclusions and perspectives

4 Exascale perspectives

Matthias Gonzalez (Univ. Paris Diderot) ASTROSIM Conference October 9, 2018 2 / 17

Star formation : context

Interstellar medium life cycle

Open questions

angular momentum transport

disc formation

fragmentation IMF/CMF


Pre-stellar phase of star formationpre-stellar dense core : R ⇠ 0.1 pcfirst Larson core : R ⇠ 10 AUsecond Larson core : R ⇠ 0.01 AU

Vaytet et al. 2013


Why studying the high-mass stars ?

high-mass stars (M>8M�, L>103 L�) :

a few (' 1%)

but dominant in energetic budgetI kinetic : outflows, jets, winds, SN explosionI radiative : luminosity, ionisation, radiative pressure

Few observational constraints

short lifetime

fewer in number

Main di↵erence compared to low-mass star formation :

still accreting when star forms

) important feedback : radiative pressure, ionization


The radiation barrier

stars up to 150 M� are observed (Figer 05,

Crowther+10 ;16)

1D analytical (Larson & Starrfield 71) and numerical(Kuiper+10) estimate of 20 M�

2D e↵ect : disc-accretion, flashlight e↵ect (Yorke &

Sonnhalter 02, Kuiper+10)

3D simulations : Rayleigh-Taylor instabilities(Krumholz+07, Rosen+16)

Yorke & Sonnhalter 02

Krumholz+09

But the magnetic field is neglected


The fragmentation issue

2 scenariicompetitive accretion (Bonnell et al. 2004)

core accretion (McKee & Tan 2003)

Interplay between radiative feedbackand magnetic field reduces thefragmentation (Commercon+11, Myers+13)

We build up on these results of isolated massive core by includingnon-ideal MHD and radiative transfer


Star formation simulation setup

We use RAMSES (Teyssier 2002) with

grey FLD (Commercon+11, Gonzalez+15 )

sink particles (Bleuler & Teyssier 2014) with protostellar feedback (Hosokawa+10)

hydro or ideal MHD or ambipolar di↵usion (Masson+12,16)

@t⇢ + r · [⇢u] = 0

@t⇢u + r · [⇢u ⌦ u + P I] = �⇢r� � �rEr + (r ⇥ B) ⇥ B

@tET + r · [u (ET + PT) � B(B · u) � EAD ⇥ B] = �⇢u · r� � Prr : u � �urEr + r ·⇣

c��R

rEr

⌘

@tEr + r · [uEr] = �Prr : u + r ·⇣

c��R

rEr

⌘+ �P⇢c(aRT 4 � Er)

@tB � r ⇥ (u ⇥ B) � r ⇥ EAD = 0

Ambipolar di↵usion EMF :


Star formation simulation setup

We use RAMSES (Teyssier 2002) with

grey FLD (Commercon+11, Gonzalez+15 )

sink particles (Bleuler & Teyssier 2014) with protostellar feedback (Hosokawa+10)

hydro or ideal MHD or ambipolar di↵usion (Masson+12,16)

Initial conditions cf. Krumholz et al. 2009, Kuiper et al. 2010

isolated massive dense core

20 K, 100 M�, R0 = 0.2 pc

finest resolution of 5 AU

B ' 170 µG (µ = 2)

x

B

solid body rotation


Morphologies with HYDRO

Results

binary system (25 M� and 14 M�)

outflow of ' 1500 AU


Morphologies with Ambipolar Di↵usion

Resultssingle star

massive and large (80 000 AU) outflow


Outflow originideal MHD

log(f

rad

/fgrav

)lo

g(f

Lor

/frad

)

Radiative origin

Ambipolar Di↵usion

Magnetic originMatthias Gonzalez (Univ. Paris Diderot) ASTROSIM Conference October 9, 2018 12 / 17

Disc propertiesDisc selection criteria (Joos et al. 2012) : ⇢ > 10�15 g/cm3, v� > 2v

r ,z , ⇢v2� > P

Hydro ideal MHD

large discdominated by P

mag

Ambipolar Di↵usion

small discdominated by P

th


Disc size

Comparison to an analytical model (Hennebelle+16)


Magnetisation

iMHD AD

AD reduces Bmax

by one order of magnitude with a plateau at about 0.3 G


Summary and perspectives

100 M� dense core collapse with AD article in prep.

magnetic outflow collimated by toroidal magnetic field

no radiative Rayleigh-Taylor instabilities

thin and small discs dominated by thermal pressure

Perspectives

grey vs multigroup/irradiation model (Kuiper et al. 2010) cf. R. Mignon-Risse

global simulation of molecular cloud collapse with turbulence

synthetic observations


Numerical perspectives : towards exascale computing

heterogeneous hardware : CPU, GPU, MIC...

more complex parallelism

load balancing, scaling (up to 100,000 cores)

fault-tolerant

I/O

=) need to adapt/re-write our codes : RAMSES + canoP (MDLS)=) development/test/maintain (e.g. RAMSES ISM)

Specificity of radiation (M)HD

implicit solver

iterative method with large matrix inversion

=) coupling with linear algebra libraries ? (Kokkos/Trilinos, PETSc)


Matthias Gonz´alez Benoˆıt Commer¸con, Neil Vaytet Rapha ...

Documents

Transcript of Matthias Gonz´alez Benoˆıt Commer¸con, Neil Vaytet Rapha ...